# How to process images taken at different Flight Heights / GSD? - PIX4Dmapper

Follow

## Description

Altitude or distance from the ground is the primary variable that affects the ground sampling distance (GSD) of the images. The GSD in the overlapping areas between the missions should not differ by more than a factor of two. So for the same project, captured with the same camera, the highest flight height at which images are taken should not exceed two times the lowest flight height:

GSD1 ≤ 2 * GSD2

(Fr * Iw) / (Sw * H1 * 100) ≤ 2 * (Fr * Iw) / (Sw * H2 * 100)

H1 ≤ 2 * H2

Where:

• GSD = Ground Sampling Distance [cm/pixel].
• Sw = sensor width [mm].
• H = flight height [m].
• Fr = real focal length [mm].
• Iw = image width [pixel].

It is usually recommended to process images captured at the same flight height, as they have the same Ground Sampling Distance (GSD). It means that all images will have the same level of detail. This facilitates the matching of keypoints between images and therefore, helps the reconstruction.

For more information on how to capture the images for terrain with height variations: Image acquisition plan for terrain with height variations.

43 out of 48 found this helpful

### Article feedback (for troubleshooting, post here instead)

• Blaž (Pix4D)

Hi Christine,

Thank you for letting us know, we corrected the equation.

Best,

• Jorge Rodriguez

In relation to this observation. How does Pix4D generate keypoints from features from two overlap photos with a different GSD?

Considering air photo of a car, if in one image the car is in a single pixel of the photo, and another overlapping photo has the same car split into 4 pixels or more, will Pix4D match the corners of the car? (considering the car perfectly square).

Also, if the corners of the pixels do not match where are the key features generated? and following this question, where are the key points located in a pixel. at the center, edges, corners, all?

Thanks

• Blaž (Pix4D)

Hi Jorge,

when it comes to images with different GDP our feature detection algorithm can cope with different scales. However, the difference in the scale should not be too big.

Best,